Hybrid Machine Learning Algorithm for Intrusion Detection Systems

Intrusion Detection Systems (IDS) rely heavily on accurate classification of network traffic to identify malicious activity. This paper explores an ensemble learning approach for IDS that prioritizes high recall to minimize false negatives (undetected malicious packets). The proposed method combines Random Forest, Gradient Boosting, and Multi-Layer Perceptron classifiers, leveraging their strengths to achieve a robust and generalizable model. The ensemble demonstrates high overall accuracy (0.9998) with excellent precision and recall for both normal and potentially malicious traffic. We further discuss the tunability of the model for even higher recall through hyperparameter optimization and class weighting. This focus on high recall is crucial for IDS, where even a single missed malicious packet can have severe security consequences. The proposed ensemble approach offers a promising solution for enhancing IDS performance and mitigating cyber threats.